Ginto AI Architecture

Ginto AI is built on a modular architecture that combines several powerful technologies.

Core Components

1. Web Layer - Caddy server with PHP-FPM handling HTTP requests
2. AI Layer - Multi-provider LLM integration (Groq, Cerebras, OpenAI)
3. Sandbox Layer - LXC containers providing isolated development environments
4. Cache Layer - Redis for sessions, caching, and real-time pub/sub
5. Proxy Layer - Node.js sandbox-proxy routing requests to containers

Request Flow

User Request → Caddy → PHP Router → Controller
                                        ↓
                              LLM Provider (Groq/Cerebras)
                                        ↓
                              Tool Execution (MCP)
                                        ↓
                              Sandbox (LXC Container)
                                        ↓
                              SSE Stream → User